**MAF1731B**: Microstructure and Trading Systems

**Project: EUR/USD** Behavior during Initial Jobless Claims statements

Professor: Villalobos Ramos Omar Antonio

Student: Esteban Márquez Delgado

Registry: if700637: https://github.com/EstebanMqz/MyST_Project

November 2022

Project

Project 1: Fundamental Analysis

Abstract

This document was prepared as Final Project for Microstructure and Trading Systems - MAF1731B class during the Autumn 2022 course at ITESO for the Bachelor of Financial Engineering.

It has the following sections:

  1. Financial Aspects.

  2. Statistic Aspects.

  3. Computational Aspects.

  4. Optimización y Backtest.

0. Libraries and Dependencies:

Note: To run this notebook it is necessary to have the following libraries contained in the requirements.txt script of this project.

To install the libraries in the script run the following cell:

In [4]:
!pip install -r requirements.txt

WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -inance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -ywinpty (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -umpy (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -finance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution - (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -arl (c:\programdata\anaconda3\lib\site-packages)
ERROR: Invalid requirement: '"""' (from line 1 of requirements.txt)
WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -inance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -ywinpty (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -umpy (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -finance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution - (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -arl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -inance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -ywinpty (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -umpy (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -finance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution - (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -arl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -inance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -ywinpty (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -umpy (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -rl (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -finance (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution - (c:\programdata\anaconda3\lib\site-packages)
WARNING: Ignoring invalid distribution -arl (c:\programdata\anaconda3\lib\site-packages)

If you prefer, manual installation can be done with the individual installation of the following libraries:

  • pandas >= 1.3.4
  • pandas_datareader >= 0.10.0
  • numpy >= 1.19.1
  • jupyter >= 1.0.0
  • chart_studio >= 1.1
  • plotly >= 5.60
  • scipy >= 1.7.3
  • regex >= 2022.3.15
  • time >= 3.11
  • datetime >= 4.7
  • MetaTrader5 >= 5.0.43
  • pytz >= 2021.3
  • fire >= 0.4.0
  • kaleido >= 0.2.1
  • Pillow >= 9.3
  • statsmodels >= 0.13.5
  • scipy >= 1.9.3
  • scikit-learn >= 1.0.2

0.1 Introduction:

Description:

Jobless claims reports are provided on a weekly basis each Thursday at 8.30 am ET by the Labour Department Employment and Training Administration and it measures Unemployment Insurance (UI) benefits in the form of weekly payments to eligible workers.

The following are the categories for jobless claims:

  1. Initial Jobless Claims: The report is a measure that declares how many citizens filed for state unemployment insurance for the first time in the previous week.

  2. Continuing Jobless Claims: It is a measure that states the number of individuals who are already receveing unemployment benefits and they re-applied for them in the previous week.

It provides Actual, Previous and Consensus estimates in (thousands) and it is an important indicator because there's an inverse relationship between jobless claims results and employment:

  • As claims decrease, higher consumption and stronger economic growth follows.
  • As claims rise, employment numbers fall, consumption decreases and economic growth declines.

Readings:

There are two types of readings for reports:

  • Beat: Lower claims numbers than the market Consensus, better than expected.
  • Miss: Higher claims numbers than the market Consensus, worse than expected.
General help:¶

In order, to retrieve specific help from the scripts tree in the project execute the help function on the script or in specific modules from the following:

  • data.py
  • functions.py
  • visualizations.py
In [1]:
## Libraries
import warnings
import pandas as pd # Analysis and data management
import pandas_datareader as pdr
import pandas_datareader.data as web #Data extraction from the web
import numpy as np
from datetime import datetime, timedelta, date
import datetime

#Plotly
import plotly.graph_objects as go 
import plotly.express as px
from plotly.subplots import make_subplots 
from scipy.optimize import minimize # Optimization 
from collections import Counter #Counter

import time
import re
import io 
import glob
import os
from os import path
import fire

import kaleido #Plotly .png (html display)
import MetaTrader5 as mt5 #Mt5 lib
import pytz #Timezone module
from PIL import Image #jpg/png import

#Statistic libs
from statsmodels.tsa.stattools import pacf 
from statsmodels.tsa.stattools import acf
import statsmodels.api as sm #qq plot
from sklearn.model_selection import train_test_split
import pylab
from statsmodels.tsa.seasonal import seasonal_decompose
from statsmodels.tsa.stattools import adfuller
from scipy.stats import kstest
%matplotlib inline

#.Py archives in repo
import functions as fn 
import visualizations as vs
import data as dt

c:\ProgramData\Anaconda3\lib\site-packages\scipy\__init__.py:146: UserWarning: A NumPy version >=1.16.5 and <1.23.0 is required for this version of SciPy (detected version 1.23.5
  warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"

1. Financial aspects

Economic Indicator
Initial Jobless Claims

In [2]:
help(dt.read_indicator)

Help on function read_indicator in module data:

read_indicator(indicator, date)
    Function that reads a .csv or .txt file contaning data 
    of economic indicators Actual, Previous and Consensus estimates.
    
    Parameters
    ----------
    + indicator: Economic indicator contained in files with Datetime,
    Actual, Previous and Consensus estimates va
    lues in cols (string).
    + date: start_date for data extraction of given indicator ('YYYY-MM-DD').
    -------
    Returns: Historic data with Datetime, Actual, Previous and Consensus 
    estimates values in cols. (dataframe).
In [3]:
date = '2018-01-01'
data_raw=dt.read_indicator('Initial_Jobless_Claims_US.txt', date)
data_raw.head()
Out[3]:
DateTime Actual Consensus Previous
0 2020-02-06 13:30:00 202.0 215.0 216.0
1 2020-01-30 13:30:00 216.0 215.0 211.0
2 2020-01-23 13:30:00 211.0 215.0 204.0
3 2020-01-16 13:30:00 204.0 216.0 214.0
4 2020-01-09 13:30:00 214.0 220.0 222.0
In [4]:
data_raw.tail()
Out[4]:
DateTime Actual Consensus Previous
105 2018-02-01 13:30:00 230.0 238.0 233.0
106 2018-01-25 13:30:00 233.0 240.0 220.0
107 2018-01-18 13:30:00 220.0 250.0 261.0
108 2018-01-11 13:30:00 261.0 245.0 250.0
109 2018-01-04 13:30:00 250.0 240.0 245.0
In [5]:
help(vs.plotly_graph2)

Help on function plotly_graph2 in module visualizations:

plotly_graph2(x, y1, y2, name1, name2, x_label, y_label, title)
    Function that plots a two-traced line+marker graph with plotly for Actual values and 
    Consensus estimates from time-series weekly statements reports.
    
        Parameters
        ----------
        x: Datetime values should be set as index for plotly chart. 
        y1: Values of reports results (Ac, Pr, Est) in df cols. 
        y2: Values of reports results (Ac, Pr, Est) in df cols.  
        name1: Label of y1 trace (str). 
        name2: Label of y2 trace (str). 
        x_label: xlabel for plot (str). 
        y_label: y_label for plot (str).          
        title: Title of the plot (str). 
    
        Returns
        -------
        Returns 2 traces (y1,y2) in a didactic graph with plotly for x.
In [6]:
vs.plotly_graph2(data_raw['DateTime'], data_raw['Actual'], data_raw['Consensus'], data_raw.columns[1],
data_raw.columns[2], data_raw.columns[0], "Wk. thousands", "Initial Jobless Claims")
Out[6]:
(None, None)

It can be assumed that Beat and Missed targets should be detected by calculating the difference between Actual and Consensus.

In [7]:
data_raw['Diff']=data_raw['Actual']-data_raw['Consensus']
data_raw.head()
Out[7]:
DateTime Actual Consensus Previous Diff
0 2020-02-06 13:30:00 202.0 215.0 216.0 -13.0
1 2020-01-30 13:30:00 216.0 215.0 211.0 1.0
2 2020-01-23 13:30:00 211.0 215.0 204.0 -4.0
3 2020-01-16 13:30:00 204.0 216.0 214.0 -12.0
4 2020-01-09 13:30:00 214.0 220.0 222.0 -6.0
In [8]:
help(vs.plotly_graph1)

Help on function plotly_graph1 in module visualizations:

plotly_graph1(x, y, name, x_label, y_label, title)
    Function that plots a one-trace line+marker graph with plotly for time-series 
    reports Differences between Actual and Consensus in weekly statements.
    
        Parameters
        ----------
        x: Datetime values should be set as index for plotly chart. 
        y: Values of reports results Difference (Actual-Consensus) in df cols. 
        name: Label of y trace (str). 
        x_label: xlabel for plot (str). 
        y_label: y_label for plot (str).          
        title: Title of the plot (str). 
        Returns
        -------
        Returns 1 trace equal to y = difference (A-C) showing if forecasts are Beaten or Missed.

  • Beat (Actual<Consensus): Economic Growth expected.

  • Miss (Actual>Consensus): Economic Decline expected.

This indicator shouldn't be taken lightly as it can be identified rather quickly how Initial Jobless Claims indicator seems to sound the alarm before economic declines.
In this 2-year sample it can be quickly noticed that the biggest difference and greater miss from estimates was in Dic 2019.
On Thursday's report, official data showed that 39,000 more americans than expected, filed for unemployment insurance for the first time during the previous week.

Just a few months later, Covid crisis made itself clear for the American economy and sooner or later, for the rest of the world.

Validations

Note: In order to convert New York timezone (EST) to GMT (MetaTrader5), validations from reports are analyzed at 14:30pm GMT for charts.

The Jobless Claims targets that were missed by the most, were registered on the following dates:

In [9]:
data_raw.sort_values('Diff').tail(10)
Out[9]:
DateTime Actual Consensus Previous Diff
52 2019-02-07 13:30:00 234.0 221.0 253.0 13.0
89 2018-05-24 12:30:00 234.0 220.0 222.0 14.0
51 2019-02-14 13:30:00 239.0 225.0 234.0 14.0
62 2018-11-29 13:30:00 234.0 220.0 224.0 14.0
40 2019-05-02 12:30:00 230.0 215.0 230.0 15.0
108 2018-01-11 13:30:00 261.0 245.0 250.0 16.0
96 2018-04-05 12:30:00 242.0 225.0 215.0 17.0
41 2019-04-25 12:30:00 230.0 200.0 192.0 30.0
53 2019-01-31 13:30:00 253.0 215.0 199.0 38.0
8 2019-12-12 13:30:00 252.0 213.0 203.0 39.0

As actual values for Initial Jobless Claims were higher than expected because the target was missed, unemployment increased and the US economy was shrinking.
The economic consequences are many, one of them can be seen in the short term in exchange rates.
In this project, EURUSD is going to be analyzed:

  • The Euro will appreciate and the USD will depreciate. This should be observed as EURUSD increases.
In [10]:
Miss1 = Image.open('Miss1.jpg')
Miss1
Out[10]:
In [11]:
Miss2 = Image.open('Miss2.jpg')
Miss2
Out[11]:

The Jobless Claims targets that were beaten by the most, were registered on the following dates:

In [12]:
data_raw.sort_values('Diff').head(10)
Out[12]:
DateTime Actual Consensus Previous Diff
107 2018-01-18 13:30:00 220.0 250.0 261.0 -30.0
54 2019-01-24 13:30:00 199.0 220.0 213.0 -21.0
93 2018-04-26 12:30:00 209.0 230.0 232.0 -21.0
60 2018-12-13 13:30:00 206.0 225.0 231.0 -19.0
101 2018-03-01 13:30:00 210.0 226.0 222.0 -16.0
97 2018-03-29 12:30:00 215.0 230.0 229.0 -15.0
43 2019-04-11 12:30:00 196.0 211.0 202.0 -15.0
92 2018-05-03 12:30:00 211.0 225.0 209.0 -14.0
45 2019-03-28 12:30:00 211.0 225.0 221.0 -14.0
44 2019-04-04 12:30:00 202.0 216.0 211.0 -14.0

With lower Jobless Claims than expected, Employment increases and the US economy has more growth than expected.

  • The USD will appreciate and the Euro will depreciate. This should be observed as EURUSD exchange rates decrease.
In [13]:
Beat1 = Image.open('Beat1.jpg')
Beat1
Out[13]:
In [14]:
Beat2 = Image.open('Beat2.jpg')
Beat2
Out[14]:
In [15]:
Beat3 = Image.open('Beat3.jpg')
Beat3
Out[15]:

The candle chart of EURUSD is plotted on a Daily timeframe to illustrate how short term consequences of Initial Jobless Claims reports are immediate but they aren't meant to be long lasting if the Euro declines as well.

In [16]:
help(dt.fx_rate)

Help on function fx_rate in module data:

fx_rate(pairs, account, pw, timeframe)
    Function that downloads exchange rates data for given timeframes with MetaTrader 5 api.
    
    Parameters
    ----------
    + pairs: Exchange Rate Currencies (ej. USDMXN, EURUSD)
    + account: Valid Mt5 username credential. (64225494)
    + pw: Valid Mt5 password credential. (movz2vvi)
    + timeframe: Data download temporality: 10M, 30M, H1, H4, D1 (str). 
    -------
    Returns: Returns OHCLV, spreads and Real Volume for selected timeframes (dataframe).

EURUSD exchange rates are downloaded from MT5 with the function above:

In [17]:
pairs, account, pw, timeframe = 'EURUSD', 64225494, 'movz2vvi', '30M'
EURUSD = dt.fx_rate(pairs, account, pw, timeframe)
EURUSD.head()
Out[17]:
time open high low close tick_volume spread real_volume
0 2018-01-02 00:00:00 1.20088 1.20094 1.20007 1.20058 107 18 0
1 2018-01-02 00:30:00 1.20058 1.20153 1.20058 1.20135 292 18 0
2 2018-01-02 01:00:00 1.20118 1.20145 1.20045 1.20129 1044 18 0
3 2018-01-02 01:30:00 1.20130 1.20143 1.20087 1.20098 630 18 0
4 2018-01-02 02:00:00 1.20087 1.20159 1.20083 1.20137 277 16 0
In [18]:
help(vs.OHCLV_csticks)

Help on function OHCLV_csticks in module visualizations:

OHCLV_csticks(fx_rates, title_1, title_2, n)
    Function that plots candlesticks, BB and MA for the given stock/forex with given temporality in dataframe (fx_rates).
    
        Parameters
        ----------
        fx_rates: fx_rates with OHLCV (dataframe).
        title_1: Upper subplot title (str). 
        title_2: Lower subplot title (str).
        n: MA & Std. ~ Window (default = 30)
    
        Returns
        -------
        Returns candlesticks time-series with Moving Average (n ~ window), Bollinger Bands (n ~ Window) and Volumes.
In [19]:
data_raw.sort_values('Diff').tail(1)
Out[19]:
DateTime Actual Consensus Previous Diff
8 2019-12-12 13:30:00 252.0 213.0 203.0 39.0

To illustrate the highest missed target:
In 2019-12-12 the EURUSD closed at 1.09748.

Immediate effects caused the EURUSD to rise and by Jan 8 price levels of 1.11 were being observed.
They weren't meant to be long lasting as by Febraury 3 (2 months later) it started to fall abruptly.

  • Technical Analysis show how by the end of the sample the EURUSD recovered by more than just a correction
    but powered as well by an upward trend because the SMA ~ 30 (d) resistance was broke.
    This pointed to a recovery for the Euro against the dollar which did in fact occur, it lasted for more than a year.
In [20]:
EURUSD_D = dt.fx_rate(pairs, account, pw, 'D1')
vs.OHCLV_csticks(EURUSD_D, 'EURUSD Chart', 'Volume', 30)
Out[20]:
(None, None)

2. Statistical aspects

1. Component of Autocorrelation and Partial Autocorrelation.

In [21]:
help(vs.create_corr_plot)

Help on function create_corr_plot in module visualizations:

create_corr_plot(index, plot_pacf=False)
    Function that graphs lines+marker AutoCorrelation and Partial AutoCorrelation 
    plot intended to model economic index Actual values.
    
        Parameters
        ----------
        index: Actual values from economic index (col) 
    
        Returns
        -------
        lines+marker AutoCorrelation and Partial AutoCorrelation 
        plots in a didactic graph with plotly.

From the next acf and pacf plots it can be concluded that there is a strong correlation at lag = 1 from Actual values.

In [22]:
vs.create_corr_plot(data_raw.Actual, plot_pacf=False),
vs.create_corr_plot(data_raw.Actual, plot_pacf=True)

2. Heteroscedasticity test.

The QQ plot shows that index data Actual values, do not form a straight line with its quantiles.
Therefore, it can't be concluded that samples come from the same distribution.

In [23]:
vs.qq(data_raw.Actual)

3. Normality test.

Suplementary to graphical assestment of normality.
The p-value is less than .05 in the Kolmogorov-Smirnov test, therefore we can reject the null hypothesis.
There's enough evidence to say Actual values don't come from a normal distribution.

In [24]:
kstest(data_raw.Actual, 'norm')
Out[24]:
KstestResult(statistic=1.0, pvalue=0.0)

4. Stationarity and Seasonality

The p-value obtained is lower than significance level of 0.05 so we can reject the null hypothesis and assume stationarity in the time-series.

In [25]:
vs.Stationarity(data_raw['DateTime'], data_raw['Actual'], 52)
Out[25]:
('p-value:', 0.027928715879526382)

6. Atypical Detection.

It can be observed from Actual Values that there are several outliers on the upside which just points to major unemployment on rare circumstances.

In [26]:
vs.Box(data_raw, data_raw['Actual'], "Actual")
Out[26]:
(None, None)

3. Computational Aspects

We are going to classify case scenarios for Actual values in relation to Consensus estimates and Previous report values as well.
This will help us determine the type of operation that should be made, given the case.

In [27]:
Cases = pd.DataFrame({'Cases':['A', 'B', 'C', 'D'],
'Assignation': ['Actual >= Consensus >= Previous',
'Actual >= Consensus < Previous ',
'Actual < Consensus >= Previous',
'Actual <= Consensus >= Previous']})

Cases
Out[27]:
Cases Assignation
0 A Actual >= Consensus >= Previous
1 B Actual >= Consensus < Previous
2 C Actual < Consensus >= Previous
3 D Actual <= Consensus >= Previous
In [28]:
vs.plotly_graph1(data_raw['DateTime'], data_raw['Diff'], data_raw.columns[4], data_raw.columns[0], "Beat or Miss", 'Diff. (Actual - Consensus)')
Out[28]:
(None, None)

As it was stated on Cases dataframe:
Missed Targets make the least favourable scenarios A, B.
The corresponding response in EURUSD is to Buy = The USD depreciates.

Beats make the most favourable scenarios C, D.
The corresponding response in EURUSD is to Sell = The USD appreciates.

In [29]:
cases=fn.case_operation(data_raw)
cases.sort_values('Diff').head(10)
Out[29]:
DateTime Actual Consensus Previous Diff Case EURUSD
107 2018-01-18 13:30:00 220.0 250.0 261.0 -30.0 D Sell
54 2019-01-24 13:30:00 199.0 220.0 213.0 -21.0 C Sell
93 2018-04-26 12:30:00 209.0 230.0 232.0 -21.0 D Sell
60 2018-12-13 13:30:00 206.0 225.0 231.0 -19.0 D Sell
101 2018-03-01 13:30:00 210.0 226.0 222.0 -16.0 C Sell
97 2018-03-29 12:30:00 215.0 230.0 229.0 -15.0 C Sell
43 2019-04-11 12:30:00 196.0 211.0 202.0 -15.0 C Sell
92 2018-05-03 12:30:00 211.0 225.0 209.0 -14.0 C Sell
45 2019-03-28 12:30:00 211.0 225.0 221.0 -14.0 C Sell
44 2019-04-04 12:30:00 202.0 216.0 211.0 -14.0 C Sell
In [30]:
cases.sort_values('Diff').tail(10)
Out[30]:
DateTime Actual Consensus Previous Diff Case EURUSD
52 2019-02-07 13:30:00 234.0 221.0 253.0 13.0 B Buy
89 2018-05-24 12:30:00 234.0 220.0 222.0 14.0 B Buy
51 2019-02-14 13:30:00 239.0 225.0 234.0 14.0 B Buy
62 2018-11-29 13:30:00 234.0 220.0 224.0 14.0 B Buy
40 2019-05-02 12:30:00 230.0 215.0 230.0 15.0 B Buy
108 2018-01-11 13:30:00 261.0 245.0 250.0 16.0 B Buy
96 2018-04-05 12:30:00 242.0 225.0 215.0 17.0 A Buy
41 2019-04-25 12:30:00 230.0 200.0 192.0 30.0 A Buy
53 2019-01-31 13:30:00 253.0 215.0 199.0 38.0 A Buy
8 2019-12-12 13:30:00 252.0 213.0 203.0 39.0 A Buy

Resulting Case Scenarios are illustrated in the following histogram:

In [31]:
vs.indicator_scenarios(cases, 'Case Scenarios')
Out[31]:
Counter
EURUSD
Sell 61
Buy 49

4. Backtest and Optimization:

In [32]:
pairs, account, pw, timeframe = 'EURUSD', 64225494, 'movz2vvi', '30M'
DateTime, Actual, Consensus, Previous, Case, Direction, pip_up, pip_down, volatility = ([] for i in range(9))
EURUSD = dt.fx_rate(pairs, account, pw, timeframe)
In [34]:
df_escenarios = fn.Optimization(cases, EURUSD)
df_escenarios.head(10)
Out[34]:
DateTime Case Direction pip_up pip_down volatility
0 2018-01-04 13:30:00 B 1 58.0 47.0 13.0
1 2018-01-11 13:30:00 B 1 91.0 88.0 1.0
2 2018-01-18 13:30:00 D -1 34.0 23.0 9.0
3 2018-01-25 13:30:00 C -1 26.0 30.0 7.0
4 2018-02-01 13:30:00 C -1 85.0 81.0 0.0
5 2018-02-08 13:30:00 C -1 14.0 12.0 3.0
6 2018-02-15 13:30:00 A 1 66.0 86.0 3.0
7 2018-02-22 13:30:00 C -1 27.0 18.0 3.0
8 2018-03-01 13:30:00 C -1 71.0 76.0 2.0
9 2018-03-08 13:30:00 A 1 108.0 97.0 2.0
In [35]:
df_escenarios.tail(10)
Out[35]:
DateTime Case Direction pip_up pip_down volatility
100 2019-12-05 13:30:00 C -1 29.0 29.0 2.0
101 2019-12-12 13:30:00 A 1 40.0 37.0 1.0
102 2019-12-19 13:30:00 B 1 4.0 1.0 1.0
103 2019-12-26 13:30:00 D -1 39.0 43.0 0.0
104 2020-01-02 13:30:00 C -1 62.0 58.0 3.0
105 2020-01-09 13:30:00 D -1 8.0 3.0 0.0
106 2020-01-16 13:30:00 C -1 11.0 12.0 1.0
107 2020-01-23 13:30:00 C -1 40.0 37.0 2.0
108 2020-01-30 13:30:00 A 1 22.0 16.0 3.0
109 2020-02-06 13:30:00 D -1 14.0 15.0 2.0
In [36]:
train, test = train_test_split(df_escenarios, test_size=0.5)
train = train.sort_values('DateTime')
test = test.sort_values('DateTime')

5. Bibliography:

  • Villalobos, O, 2021. Python Project Template. Available at: https://github.com/OmarVillalobos/python-project.

  • ETA, (2022). United States Department of Labor. Available at: https://www.dol.gov/agencies/eta (Accessed: November 21, 2022).